Resistive Switching Random Access Memory Research Articles

Extremely Low Program Current Memory Based on Self-Assembled All-Inorganic Perovskite Single Crystals.

Memory devices based on lead halide perovskite have attracted great interests because of their unique current-voltage hysteresis. However, current memory devices based on polycrystalline perovskites usually suffer from large intrinsic electronic current and parasitic leakage current due to the existence of grain boundaries, which further leads to high power consumption. Here, a low-power resistance switching random-access memory device is demonstrated by assembling single-crystalline CsPbBr3 on Ag electrodes. The assembled structure serves as a bipolar nonvolatile resistance switching memory device with a low program current (∼10 nA), good endurance, long data retention (>103 S), and big on/off ratio of ∼103. The low program current results in a power of ∼3 × 10-8 W, which is much lower than that of polycrystalline perovskite-based devices (10-1-10-6 W). It is found that the formation and annihilation of Ag and bromide vacancy conductive filaments contribute to the significant resistive switching effect. At a low resistive state, the conductive filaments originate from the accumulation of Br- ions at the drain. Furthermore, the conductive filaments are proved to be a cone shape, shrinking from the drain to the source.

Stretchable and Wearable Resistive Switching Random‐Access Memory

In the era of big data, with the development and application of 5G technology, artificial intelligence technology, and wearable electronics, the acquisition, storage, search, sharing, analysis, and even visual presentation require huge amounts of data in storage and processing at any time. This propounds a strong demand for the stretchable and wearable memory devices with ultrafast switching, high density, and high energy efficiency. Herein, an overview of the advanced progresses made on developing stretchable resistive switching random‐access memory (RRAM) is presented, concentrating on the fundamental components, materials selection principles, and configuration design strategies. The basic electrical and mechanical properties of stretchable components to meet the functionality of memory storage are summarized. Novel materials with the intrinsically stretchable features toward the realization of stretchable RRAM are discussed. Ingenious configuration design strategies to obtain stretchable RRAM are evaluated. Finally, the challenges and perspectives for the practical applications of stretchable RRAM are also addressed. Some helpful guidelines are provided in promoting the development on stretchable and wearable RRAM for making deployable and portable artificial intelligence and smart devices.

LRS retention fail based on joule heating effect in InGaZnO resistive-switching random access memory

This study reports the low-resistance state retention fail of InGaZnO resistive-switching random access memory (ReRAM) under constant DC bias stress conditions by Joule heating effect. There were the abrupt state changes of InGaZnO ReRAM devices with high voltage stress over 0.6 V because of thermal energy in conducting filament. In addition, SPICE simulation was conducted with verilog-A to verify this retention fail mechanism. We believe these results are potentially useful to the analysis on the retention fail properties of ReRAM devices as well as the system-level simulations with reliability-awareness.

Uniform multilevel switching of graphene oxide-based RRAM achieved by embedding with gold nanoparticles for image pattern recognition

Traditional metal-oxide semiconductor devices are inadequate for use in artificial neural networks (ANNs) owing to their high power consumption, complex structures, and difficult fabrication techniques. Resistive random access memory (RRAM) is a promising candidate for ANNs owing to its simple structure, low power consumption, and excellent compatibility with CMOS. Moreover, it can mimic synaptic functions because of its multilevel resistive switching (RS) behavior. Herein, we demonstrate highly uniform RS and a high on/off ratio of RRAM based on graphene oxide by embedding gold nanoparticles into the device. This allowed reliable multilevel storage. Further, multilevel RRAM based on spike-timing-dependent-plasticity learning rules was used for image pattern recognition. These findings may offer a route to develop reliable digital memristors for ANNs.

Highly stable, solution-processed quaternary oxide thin film-based resistive switching random access memory devices via global and local stoichiometric manipulation strategy

Optimization and performance enhancement of a low-cost, solution-processed InGaZnO (IGZO) resistance random access memory (ReRAM) device using the manipulation of global and local oxygen vacancy (Vo) stoichiometry in metal oxide thin films was demonstrated. Control of the overall Ga composition within the IGZO thin film reduced the excessive formation of oxygen vacancies allowing for a reproducible resistance switching mechanism. Furthermore, sophisticated local control of stoichiometric Vo is achieved using a 5 nm Ni layer at the IGZO interface to serve as an oxygen capturing layer through the formation of NiOx, consequently facilitating the formation of conductive filaments (CFs) and preventing abrupt degradation of device performance. Additionally, reducing the cell dimension of the IGZO-based ReRAMs using a cross-bar electrode structure appeared to drastically improve their performances parameters, including operating voltage and resistance distribution due to the suppression of excessive CFs formation. The optimized ReRAM devices exhibited stable unipolar resistive switching behavior with an endurance of >200 cycles, a retention time of 104 s at 85 °C and an on/off ratio greater than about 102. Therefore, our findings address the demand for low-cost memory devices with high stability and endurance for next-generation data storage technology.

Investigation of the Thermal Recovery From Reset Breakdown of a SiN x -Based RRAM

In this article, we report a method for recovering a resistive-switching random access memory (RRAM) from the reset breakdown and enhancing the endurance characteristics. A SiNx-based RRAM device has been fabricated and its switching characteristics are analyzed with a particular interest in reset breakdown. It has been found that the SiNx RRAM in permanent reset breakdown can be revived to high-resistance state (HRS) by thermal recovery, a low-temperature annealing method. The memory window could be widened even with enhanced reset stability and conductance distribution. Temperature-dependent conductance change has been measured in order to figure out the substantial component of conductive filament and the mechanism of thermal recovery. The experimental evidences show that the reset breakdown process is the result of unwanted Si dangling bond (Si-DB) formation. By thermal recovery, Ni filament could be ruptured without applying a high electric field which induces negative set.

Observing topotactic phase transformation and resistive switching behaviors in low power SrCoOx memristor

Abstract Recently, complex oxides have been shown to be promising candidate in dielectric materials of resistive random access memory (RRAM). However, the detailed switching information of complex oxide RRAM is still insufficient, and direct observation of the whole switching process is required to figure out the mechanism. In this study, we deposited SrCoOx (SCO) on a niobium-doped SrTiO3 substrate as the dielectric layer via pulsed laser deposition (PLD). The novel SCO device possesses excellent RRAM properties, high cycling endurance, a long data retention time, and uniform distributions of the high resistance state (HRS) and low resistance state (LRS) resistance and Set/Reset voltage. Furthermore, the switching mechanism was investigated by using transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM), which showed that the switching behavior resulted from the topotactic phase transformation. In addition, the whole switching process was observed through in situ TEM, and the results strengthened the findings of the ex situ experiment. The discussion of this switching behavior provides support for a novel aspect of the RRAM switching mechanism and also a new option for the dielectric material in RRAM.

Forming-ready resistance random access memory using randomly pre-grown conducting filaments via pre-forming

The electrical forming is among the issues to be considered during practical applications of resistive switching random access memory (ReRAM) arrays. The typical electrical forming process increases the power budget and circuit complexity of the ReRAM arrays. From the material engineering point of view, most competitive ReRAM materials require higher forming voltages than the set voltages for efficient device operation. Considering these bottlenecks, herein, we report a novel method for a substantial reduction of the forming voltage to a range close to the set voltage. The proposed forming method covers cells in more than the 700 μm range, using a one-time pre-forming process. Randomly grown filaments completely formed over the active layers during pre-forming are reused for device operation without the need for further forming. To validate this method, we fabricated 8 × 8 ReRAM arrays with two conductive filament-based mechanisms namely valance change and electrochemical migration and each of the eight cells in line was set as the test sample to confirm the completion of the forming. All the eight cells in line had initial set voltages of ~1 V, while the reference samples had forming voltages of ~3 V. The results indicate that electrical forming had already occurred in the eight cells under the one-time bias application; thus, a range of cells spread over more than 700 μm was formed with the proposed method. This method is advantageous for the circuit design of ReRAM arrays with forming-free behavior.

An Improved RRAM-Based Binarized Neural Network With High Variation-Tolerated Forward/Backward Propagation Module

Binarized neural network (BNN) enables resistive switching random access memory (RRAM) with high nonlinearity and nonsymmetry to realize online training, using an RRAM & comparator structure. In this work, a new hardware implementation approach is proposed to improve the efficiency of BNN. In the approach, an 1T1R array-based propagation module is introduced and designed to realize the computing acceleration of fully parallel vector-matrix multiplication (VMM) in both forward and backward propagations. Using the 1T1R-based propagation module, high computing efficiency is achieved in both training and inference tasks, improving by 50× and 177×, respectively. To solve the computation error caused by device variation, a novel operation scheme with low gate voltage is proposed. With the operation scheme, the RRAM variation is dramatically suppressed by 74.8% for cycleto-cycle and 59.9% for device-to-device. It enables high-accuracy VMM calculation and, therefore, achieves 94.7% accuracywith a typical BNN, showing only 0.7% degradation from the ideal variation-free case.

Uncovering the Indium Filament Revolution in Transparent Bipolar ITO/SiOx/ITO Resistive Switching Memories

Transparent resistive switching random access memory (ReRAM) is of interest for the future integrated invisible circuitry. However, poor understanding of its working mechanism in transparent ReRAMs with the indium tin oxide (ITO) electrode is still a critical problem and will hinder its widespread applications. To reveal the actual working mechanism in transparent ReRAMs with the ITO electrode, we investigate the transparent ITO/SiOx/ITO memory devices (∼82% transmittance in the visible region) and compare it with ITO/SiOx/Au memory devices, which both can exhibit reproducible bipolar switching. The indium (In) filament evolution, which accounts for the bipolar switching behaviors in the ITO/SiOx/ITO (or Au) memories, is directly observed using transmission electron microscopy on samples with different memory states (electroformed, ON, and OFF). These studies uncover the microscopic mechanism behind the bipolar switching in SiOx-based ReRAM devices with the ITO electrode, providing a general guidance for the design of high-performance ReRAMs with large scalability and high endurance.

A Study on the Interfacial Structure Design of Self‐Compliance Ni/HfOx/Ni Bipolar Selectors and Trap‐to‐Trap Tunneling Mechanism Caused by Redox Reaction

Selectors are considered some of the most significant devices applied to the integration of resistive random access memory (RRAM). Self‐compliance selectors promote the stability of one selector and one resistor (1S1R) cell to reduce additional compliance current circuits without compromising RRAM switching performance. Herein, a Ni/HfOx/Ni bipolar selector with special cluster microstructures is fabricated via a facial magnetron sputtering technology, which has self‐compliance (±12 μA) performance and ultrahigh voltage endurance (±10 V). To investigate the mechanism, theoretical models for the self‐compliance selectors are proposed. The results demonstrate that traps and special interfaces are formed through redox reactions. Resistive switching properties are caused by the trap‐to‐trap tunneling structure of the HfOx–NiOy layers. At the interfaces of HfOx–NiOy oxide layers, the tunneling current reaches a saturation state, leading to the self‐compliance phenomenon, which provides a theoretical and research foundation for the development of RRAM technology.

Kernel Application of the Stacked Crossbar Array Composed of Self‐Rectifying Resistive Switching Memory for Convolutional Neural Networks

Herein, a feasible method is provided for circuit implementation of the convolutional neural network (CNN) in neuromorphic hardware using the multiple layers‐stacked resistance switching random access memory (ReRAM). The specific ReRAM is accompanied by self‐rectification functionality. The single‐input multiple‐output (SIMO) scheme is an optimum method in the extraction of the features of a letter with a versatile selection of the intended features, whereas the multiple‐input single‐output (MISO) scheme provides a highly efficient method to extract the features from the color image, which is composed of several component color images. The Pt/HfO2−x/TiN‐based self‐rectification ReRAM that is integrated into the sidewalls of the two‐layer structure provides a sound framework for the circuit implementation of the SIMO and MISO schemes. The appropriate selection of the kernels for image compression and feature extraction greatly facilitates the CNN in neuromorphic hardware.

Rapid thermal annealing on the atomic layer-deposited zirconia thin film to enhance resistive switching characteristics

The resistive switching random access memory (RRAM) device has received a great interest for the next-generation non-volatile memory application, and resistive switching (RS) characteristics are mainly affected by conductive oxygen vacancies ([V··]) within switching material. Various effective approaches with materials, doping, and thermal treatments have been attempted to achieve the stable RS behaviors. Particularly, thermal annealing is considered as an efficient knob to control [V··] compared to other approaches. However, research on thermal treatment still lacks results and further research efforts are still needed to improve the RS characteristics of the devices. In this work we investigated the RS characteristics of Ti/ZrOx/Pt-structured RRAM device in comparison without and with postrapid thermal annealing (RTA) temperature range under oxygen ambient. The as-fabricated device with atomic layer-deposited ZrOx switching layer exhibited conducting characteristics, which is related to a relatively high amount of [V··] within switching medium. It indicates that moderate amount of [V··] apparently determines the appropriate RS behaviors. With this regard, we modulated the [V··] in ZrOx thin films by employing RTA in the ranges of 500 °C to 800 °C at the oxygen ambient for 60 s. Unlike device without RTA, we observed the stable RS characteristics from device with RTA and device annealed at 700 °C exhibits the excellent bipolar RS characteristics such as higher Ron/Roff, smaller cycle-to-cycle switching variation, better endurance, and longer retention among RTA conditions, indicating moderate amount of [V··] formed within ZrOx thin film layer. Moreover, increasing ALD ZrOx thickness shows the further improvement in the RS characteristics and RTA on the thicker ZrOx device still improves the RS behaviors. This research indicates that modulating [V··] by fast thermal annealing on the ALD zirconia material can provide the proper RS characteristics of the non-volatile memory applications.

Transition from rectification to resistive-switching in Ti/MgF2/Pt memory

Magnesium fluoride is a promising candidate for resistive-switching random access memory (RRAM) with biodegradable property. However, the underlying resistive-switching (RS) mechanism and conduction mechanism in MgF2 material is still not fully studied, which limits the further performance optimization. In this work, we have demonstrated the rectification and bipolar RS behaviors in Ti/MgF2/Pt device. The un-Formed device exhibits stable rectifying characteristics, with a rectifying ratio of ∼103 at ±3 V. Furthermore, transition from rectification mode to RS behavior can be achieved by a Forming process. Then the analogue switching properties with high uniformity was obtained with quasi-DC sweeps, and gradual modulation of multi-state was further realized. At last, the conduction and switching mechanism was discussed.

(Invited) Balancing the Source and Sink of Oxygen Vacancies for the Resistive Switching Memory

The high non-uniformity and low endurance of the resistive switching random access memory (RRAM) are the two major remaining hurdles at the device level for mass production. Incremental step pulse programming (ISPP) can be a viable solution to the former problem, but the latter problem requires a material-level innovation. In valence change RRAM, electrodes have usually been regarded as inert (e.g., Pt or TiN) or oxygen vacancy (VO) sources (e.g., Ta), but different electrode materials can serve as a sink of VO. In this work, an RRAM using a 1.5nm-thick Ta2O5 switching layer is presented, where one of the electrodes was VO-supplying Ta and the other was either inert TiN or VO-sinking RuO2. While the TiN could not remove the excessive VO in the memory cell, the RuO2 absorbed the unnecessary VO. By carefully tuning (balancing) the capabilities of VO-supplying Ta and VO-sinking RuO2 electrodes, an almost invariant ISPP voltage and a greatly enhanced endurance performance can be achieved.

Volatile Resistive Switching Memory Based on Ag Ion Drift/Diffusion—Part II: Compact Modeling

Resistive-switching random access memory (RRAM) devices based on filamentary switching are attracting widespread interest for their unique properties, such as high ON–OFF ratio, ultrasteep slope, good endurance, and low-current operation. Recently, volatile RRAMs based on Ag drift and diffusion were also demonstrated for possible applications such as crosspoint array selectors and neuromorphic computing. However, the mechanism of the volatile switching, namely, the spontaneous dissolution of the Ag filament, is still not clear. A deep understanding of the metallic filament formation and spontaneous disruption would strongly help the engineering of the device for optimized performances. Here, we present a numerical physics-based drift/diffusion modeling framework to describe the threshold switching, ${I}$ – ${V}$ characteristics, and morphological evolution of the metallic filament. The model can support TCAD-type device simulations for scaling, reliability, and variability studies.

Hardware implementation of RRAM based binarized neural networks

Resistive switching random access memory (RRAM) has been explored to accelerate the computation of neural networks. RRAM with linear conductance modulation is usually required for the efficient weight updating during the online training according to the back-propagation algorithm. However, most RRAM devices usually show the nonlinear characteristic. Here, to overcome the dilemma, we designed a novel weight updating principle for binarized neural networks, which enables the nonlinear RRAM to realize the weight updating in efficiency during online training. Moreover, a vector-matrix multiplication is designed to parallel calculate the dot-products of the forward and backward propagation. 1 kb nonlinear RRAM array is fabricated to demonstrate the feasibility of the analog accumulation and the parallel vector-matrix multiplication. The results achieved in this work offer new solutions for future energy efficient neural networks.

Dependence of Switching Probability on Operation Conditions in GexSe1–xOvonic Threshold Switching Selectors

Ovonic threshold switching (OTS) selector is a promising candidate to suppress the sneak current paths in resistive switching random access memory (RRAM) arrays. Variations in the threshold voltage ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V} _{\text {th}}$ </tex-math></inline-formula> ), and the hold voltage ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V} _{\text {hd}}$ </tex-math></inline-formula> ) have been reported, but a quantitative analysis of the switching probability dependence on the OTS operation conditions is still missing. A novel characterization method is developed in this letter, and the time-to-switch- ON/OFF ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\text{t}_{ \mathrm{\scriptscriptstyle ON}}/\text{t}_{ \mathrm{\scriptscriptstyle OFF}}$ </tex-math></inline-formula> ) at a constant <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V} _{\text {OTS}}$ </tex-math></inline-formula> is found following the Weibull distribution, based on which the dependence of switching probability on pulse bias and time can be extracted and extrapolated, and the switching probability can be ensured with appropriately chosen operation conditions. The difference between square and triangle switching pulses is also explained. This provides a practical guidance for predicting the switching probability under different operation conditions and for designing reliable one-selector-one-RRAM (1S1R) arrays.

The observation of resistive switching characteristics using transparent and biocompatible Cu2+-doped salmon DNA composite thin film

For the potential switching bio-memory device application using DNA composite thin film, we fabricated and characterized the transparent and biocompatible resistive switching random access memory (RRAM) device within the structure stacking of Pt/Cu2+ doped salmon DNA/FTO, where Cu2+ doping into salmon DNA was solution-processed. The device shows good bipolar switching characteristics with SET and RESET processes at negative and positive sweeps, respectively, with switching memory window greater than 103 ratios. The device was observed to be in low resistance state as its pristine state and an initial RESET state was necessary to achieve programmable SET and RESET cycles. Based on the electrical characteristics of the Cu2+-doped salmon DNA-based RRAM device we propose a switching mechanism with the formation and rupture of conductive filaments due to the migration of Cu2+ during the electrical stress. Our understanding could contribute to the engineering of biomaterial memory switching medium for the environmentally benign, biocompatible and biodegradable memory storage devices.

Diffusion of oxygen in amorphous tantalum oxide

The switching in resistive random-access memory (RRAM) relies on the migration of ions in sputtered amorphous transition metal oxides to both create and annihilate conductive filaments. A comprehensive understanding of ionic diffusion in these materials is therefore critical for optimizing future RRAM devices. While predicting diffusion barriers in crystalline materials is straightforward, the complex energy landscape in disordered materials makes analysis of the effective diffusion barrier inherently more difficult. Using classical molecular dynamics, we examine oxygen diffusion at various temperatures in amorphous tantala $({\mathrm{Ta}}_{2}{\mathrm{O}}_{5})$, a leading material for RRAM applications. We compare the predicted activation energies for self-diffusion to those from isotope tracer diffusion experiments. We find a diffusion activation energy of 1.55--1.60 eV, which is higher than that measured in isotopic diffusion experiments $(1.2\ifmmode\pm\else\textpm\fi{}0.1\phantom{\rule{0.16em}{0ex}}\mathrm{eV})$ on amorphous ${\mathrm{Ta}}_{2}{\mathrm{O}}_{5}$. However, the predicted values are comparable to the activation energy $(1.60\ifmmode\pm\else\textpm\fi{}0.18\phantom{\rule{0.16em}{0ex}}\mathrm{eV})$ determined from secondary ion mass spectroscopy analysis of isotopic diffusion experiments of sintered L-${\mathrm{Ta}}_{2}{\mathrm{O}}_{5}$ samples. We find that this difference could be due to issues of deposited film porosity, film stoichiometry, or else a preference for the empirical potentials to form nanocrystalline regions.